I was thinking last night about how to find possible shorting candidates. Came up with the following. I'm not a math guy, so there might be easier/better ways to describe parabolic-ish behavior, but using Fibonaccis in this case should give me a nice, accelerating uptrend, perhaps nearing an unsustainable momentum. At least, it should make for a nice list of shorting candidates.

5-period Linear Regression Slope greater than 8-period Linear Regression Slope and
8-period Linear Regression Slope greater than 13-period Linear Regression Slope and
13-period Linear Regression Slope greater than 21-period Linear Regression Slope and
21-period Linear Regression Slope greater than 34-period Linear Regression Slope and
34-period Linear Regression Slope greater than 55-period Linear Regression Slope

Probably want check that all these slopes are positive, or rising, as well.

This criteria might turn out to be too restrictive. If scans don't return enough tickers I may want to start dropping off one Linear Regression Slope at a time to increase the scan hit rate. In other words, remove the 34 > 55 requirement first, rescan, then remove the 21 > 34 requirement, rescan, etc. So if there's an easy way to explain how to modify the PCF, that would be awesome. There's always a balance between too many and too few tickers to look at.

***

This morning, I found a 2012 MarketWatch article Elder co-authored about shorting Apple at the time. (http://www.marketwatch.com/story/parabolic-move-makes-apple-a-short-2012-04-12).

In the article he describes his criteria for identifying parabolic rallies. Since he's smarter than me, and since his criteria is specific and scannable, I'd like to try Elder's way also. (Good artists copy; great artists steal ;)

Here's the relevant part of the article. His examples use monthly charts, which is too long for me - I'd use it on smaller timeframes but that's another story. Can this be PCF-ified?

To receognize parabolic rallies, we use a cleary defined set a rules. A parabolic rally has three parameters:

• Two closes outside of a three standard deviation channel of a 24-month Average True Range (ATR).
• Five-month moving average is within 20% of the three standard deviation of the 24-month ATR (monthly 3-ATR channel).
• Minimum drawdown from high to close is less than 15% in the past six months.

When a market moves outside its monthly 3-ATR channel for two consecutive months, the monthly 5MA is within 20% of the 3ATR and it has not experienced a closing drawdown of more than 15% in the past six months. we consider this a parabolic move.

[End of article excerpt]

The good news about Elder's use of monthly charts is that the data is still in TC2000 and we can double-check the examples against the PCF.

I'll include his chart for his first example, and list a few of the others. If you want to see all the charts, you have to use the Internet Wayback Machine's archive copy of the article...

https://web.archive.org/web/20120414193510/http://www.marketwatch.com/story/parabolic-move-makes-apple-a-short-2012-04-12

His first example is the parabolic move in crude oil in 2008. Oil closed outside its monthly 3ATR for the second time in June 2008. Oil topped out the following month. The move back into the 3-ATR channel and subsequent break of the monthly 5-MA sent oilt crashing...

His other examples...

• Nasdaq, March 2000. In April the bubble popped.

• Silver, April 2011. In April silver closed above the monthly 3ATR for the 3rd consecutive month. The following month saw the top in silver as it moved back into the 3ATR channel. and closed below its 5MA.

• Apple March 2000. Closed outside the 3ATR channel for 6 consecutive months. Once Apple pulled back into its 3ATR channel and crossed below the 5MA it went into a severe decline. 

He lists a few more, but that should suffice.

As I typed this, I realized Elder mentions very specific criteria for when to possibly short these. A PCF of this would be awesome too to test as a short entry signal...

• Pull back inside its 3 ATR channel. (Remember he used a 24-period ATR above).

• Closed below its 5-period moving average.

At the time he wrote the article (April 2012) Elder was making the case for another Apple sell off. Once I'm done here I'll have to go back and pull up the chart and see how the above short entry signal would have played out.

***

Finally, if you can refresh my memory, how do I plot a vertical spike on the Price History pane when the above PCFs/conditions are true? I'd like to visually inspect prior examples.

Thanks for all your help, Bruce!

I just realized Elder didn't specify what kind of moving average his 5-period moving average was, but he's pretty much an EMA guy.

He considers Simple Moving Averages to be "outdated since the advent of computers". LOL. Dr. Elder is even older than I am. Love his stuff, though.

Also, the fourm kind of butchered the second link above.

The Internet Wayback Machine's archive of the story is at: http://bit.ly/2v8PSsN

The MarketWatch.com copy of the article has all the text, but only the first chart. The rest of the chart examples are MIA: http://on.mktw.net/2ua8MlI

Your linear regression slope test could be written as follows.

(FAVGC5 - AVGC5) / 4 > (FAVGC8 - AVGC8) / 7 AND (FAVGC8 - AVGC8) / 7 > (FAVGC13 - AVGC13) / 12 AND (FAVGC13 - AVGC13) / 12 > (FAVGC21 - AVGC21) / 20 AND (FAVGC21 - AVGC21) / 20 > (FAVGC34 - AVGC34) / 33 AND (FAVGC34 - AVGC34) / 33 > (FAVGC55 - AVGC55) / 54 AND FAVGC55 > AVGC55

You are almost definitely going to need to relax the requirements somewhat.

This quote bothers me, "...three standard deviation channel of a 24-month Average True Range (ATR)."

Standard deviation and average true range are two completele different things, but I am just going to assume it means 3 times the standard deviation from a moving average centerline of some sort (so Keltner Channels or something similar). I am also going to use exponential moving averages throughout.

TrueInRow(C > XAVG(C + 3 * ATR, 24), 2) = 2 AND ABS(XAVGC5 / XAVG(C + 3 * ATR, 24) - 1) <= .2 AND MIN(STOC1, 5) > 85

I read that as a channel using the 24-period ATR as the centerline, and a channel that is three times as wide as the standard deviation of the last 24 ATR values. So the upper channel line is 1.5 standard deviations above the centerline and the lower channel line is 1.5 standard deviations below.

Again, I'm a programmer, not a math or stats guy (wish I were), so we may be saying the same thing and I'm just too thick to understand :)

If not, can you code it my way, and I can try both? Again, since the data is still in TC2000 I can pull up the charts. The charts in the article are really just illustrations rather than screenshots, but they're better than nothing.

Average true range does use the same units as price (usually dollars), but it is rarely actually near price when plotted in the same scale. For most normally trading stocks, average true range is quite a bit lower than price.

It is definitely different than my interpretation but I am fairly confident that the interpretation proposed does not make any sense mathematically. That said, here you go.

TrueInRow(C > AVG(ATR,24) + 3 * SQR(ABS(SUM((ATR) ^ 2, 24)- 24 * AVG(ATR, 24) ^ 2) / 24), 2) = 2 AND ABS(XAVGC5 / (AVG(ATR,24) + 3 * SQR(ABS(SUM((ATR) ^ 2, 24) - 24 * AVG(ATR, 24) ^ 2) / 24)) - 1) <= .2 AND MIN(STOC1, 5) > 85

To give you an idea of what I mean, try plotting the following as a Custom PCF Channel Indicator.

LOL! But tell me how you really feel. I love you, Bruce!

I'm going to have to think about this until I can speak good English. I get that ATR is nowhere near price when plotted in the same scale...

Yesterday, QQQ closed at 138.66. Its ATR(24) was 1.97.

Friday, QQQ closed at 137.76, ATR(24) was 1.93.

If you go back 24 trading days, there may be some ties, but generally you'll see 24 unique values for the ATR(24) plot.

If you take that distribution of 24 values, and find the standard distribution, you should be able to draw a channel around them.

Hold on.

Actually, I just realized where my disconnect was. I was thinking of drawing a channel around ATR. Elder's article - and illustrations - plot the channel around price. 

I'm not sure how to proceed. I could email Dr. Elder; he's responded in the past, usually.

Given he's a trained psychiatrist, trader, and author rather than a CMT, mathematician, or statistician, maybe he misspoke and the editor didn't catch the issue you mentioned.

Let me pull up a chart and noodle around. Back in a few minutes.

ATR would be in the same scale as channels based on the standard deviation of ATR plotted around the average of ATR. This makes mathematical sense, but only briefly considered it because it doesn't match the pictures and I'm am pretty sure it does not make logical sense considering we are supposed to be looking for a parabolic trend in price instead of range (and I think you have already come to this same conclusion).

TrueInRow(ATR > AVG(ATR,24) + 3 * SQR(ABS(SUM((ATR) ^ 2, 24)- 24 * AVG(ATR, 24) ^ 2) / 24), 2) = 2 AND ABS(XAVG(ATR, 5) / (AVG(ATR,24) + 3 * SQR(ABS(SUM((ATR) ^ 2, 24) - 24 * AVG(ATR, 24) ^ 2) / 24)) - 1) <= .2 AND MIN(STOC1, 5) > 85

Okay, I think I got it. :)

Just read your latest response. Agreed. Mathematically, maybe. Logical? Not so much.

It's looking more like Elder's description was a little hinky and it slipped by the editor.

I went back and used his 2007 Apple example as a reference because it was very specific. Here's what he said...

Apple, nevertheless, had another parabolic move in 2007. By December of that year it closed outside of its 3-ATR channel for 8 consecutive months. 

Here's the illustration from the article...

The following month Apple moved back inside its 3ATR channel and closes below its 5MA. 

In TC2000 I tried plotting Bollinger Bands and Keltner channels around price, and BBs and KCs around ATR(24). I tried lots of stuff.

What it now looks like is Elder essentially meant a Keltner Channel around price. Below I plotted a Keltner Channel on the monthly AAPL chart. 24 periods. 3.00 ATR multiplier. EMA centerline.

My chart matches Elder's narrative, exactly. April 2007 was close, but AAPL did close just inside the channel. Then there are exactly 8 consecutive closes above the channel. Then the following month AAPL closes both inside the channel and below the EMA(5)...

I guess it's like I said earlier: never trust anything you read on the internet, except for your stuff :)

Good thing Elder included illustrations because the text doesn't seem quite right.

Okay, so I think that leaves us with the following criteria for Elder's parabolic rally:

• At least two consecutive closes outside of a Keltner Channel with 24 periods, 3.00 ATR multiplier, and EMA centerline.
• 5-period EMA is within 20% of the upper channel line.
• Minimum drawdown from high to close is less than 15% during the past 6 periods.

So, what do our PCFs look like now? Can we put them all in one place here? (I've obfuscated this post enough as it is).

Also the PCF for a short entry signal:

• Pull back inside the channel.
• Close below EMA(5).

Also, if you can refresh my memory ... how can I plot a vertical spike on the Price History pane when PCF conditions are true? Especially now that we've figured out what Elder really meant, I'd like to visually backtest some charts.

It took the better part of a day, but this makes sense now. And since the criteria seems reasonable and specific (i.e., scannable) I think letting EasyScan do its thing on any and all time frames might bring up some nice shorting opportunities.

That is pretty close to what we had.

TrueInRow(C > XAVG((H + L + C) / 3 + 3 * ATR, 24), 2) = 2 AND ABS(XAVGC5 / XAVG((H + L + C) / 3 + 3 * ATR, 24) - 1) <= .2 AND MIN(STOC1, 6) > 85

For the pullback:

C < XAVG((H + L + C) / 3 + 3 * ATR, 24) AND C < XAVGC5

If we also assume that the other formula was true during the previous bar.

C < XAVG((H + L + C) / 3 + 3 * ATR, 24) AND C < XAVGC5 AND TrueInRow(C1 > XAVG((H1 + L1 + C1) / 3 + 3 * ATR1.1, 24), 2) = 2 AND ABS(XAVGC5.1 / XAVG((H1 + L1 + C1) / 3 + 3 * ATR1.1, 24) - 1) <= .2 AND MIN(STOC1.1.1, 6) > 85

The easiest way to plot a spike when a condition is true is to just use the Condition Formula as the Boolean Formula in a Custom PCF % True Indicator.

You're awesome.

One last thing. This is pretty cool...

Elder wrote that article about shorting Apple for MarketWatch.com on 4/12/2012. Here's what he said at the time...

It took a few more months, but when AAPL eventually pulled back inside its channel and closed below its EMA(5) in October, things went south in a hurry. Keep in mind, Elder wrote the article in April, and the short entry signal did not flash until October.

Sweet...

Of course, there are no guarantees this will work the same on other timeframes, though generally speaking folks say markets are fractal. I guess we'll see.

Now that we have the PCFs, we can scan other timeframes, and begin plotting world domination. :)

One (more) last thing...

Here's how Elder ended that article. Bolding is mine:

While we were working on this article, we received a call from a friend. She got married in February and received ten shares of Apple. Her gift was worth $5,000 on the day of her wedding, but it is over $6,000 today. She wanted to sell all her stocks and buy more Apple. We told her instead to put a protective stop well above $500 per share to protect her gift. She did not know what a stop was, fairly typical of the folks who pile into a parabolic move.

These are some of the folks whose money you'll be taking with this setup. I have no problem taking money from a newlywed bride, but this is why I hesitate in getting too excited about the other timeframes. Casual traders/investors chase stocks on longer timeframes. Shorter timeframes are (usually) populated with more serious traders, and the pickings might not be so easy.

That said, we've all seen tickers overextended on all timeframes. Even algos are written by greedy people :)

I guess it's just time to scan and test and see if this setup works.

Thanks again, Bruce!!

The (third) last thing...

This kind of ties in with the price/momentum divergence thread I had with Bruce elsewhere, but around the 2:20 mark of this Vimeo video Elder shows why he shorted Apple in 2012 based on (among other things) the Price/MACD HIstogram divergence displayed on his scanner.

The cool part is at the 2:40 mark Elder mentions the MarketWatch.com article. It all comes full circle.

Moral of the story? There's more than one way to fleece a newlywed bride. :)

Um, I have a much simpler view, but looking at the monthly chart, isn't AAPL setting up the same way now? It could run up for months from here, or not, but an unsophisticated viewing tells me to put it on my calendar for a look the first weekend of each month. Thanks for the analysis and the education.

Bruce, I have one last (I think) PCF request related to this. See below, please :)

All true. Though this particular scan wouldn't classify AAPL's current monthly chart as "parabolic" (only May closed above the 3-ATR channel). An unsophisticated viewing can almost certainly identify a possible shorting opportunity here.

But, um, let me ask this ... how long would an unsophisticated viewing of every ticker on every timeframe take? 

I'm not saying these scan criteria will work on shorter timeframes, and unfortunately TC2000 doesn't include a backtesting feature. Only one way to find out.

I'm not saying Elder was a genius for shorting Apple in 2012. I'm not saying you have to be sophisticated to short parabolic stocks profitably. I'm not even saying you need an algo to short parabolic stocks. The eyeballs & brain of a decent discretionary trader will suffice.

All I'm saying is ... wouldn't it be nice to have an EasyScan continuously monitor the entire market in all timeframes, rather than manually flip through a limited number of stocks at the end of each month?

It's never as simple as running a scan, of course ... I'm just trying to leverage a concept any unsophisticated trader can spot on a chart and ask EasyScan to do the heavy lifting while I watch The Golf Channel :)

Bruce, ...

Can we mix timeframes in a PCF? I'd like to know when (1) the ticker pulls back inside its - in this case - monthly 3ATR channel and (2) closes below its EMA21 on its weekly chart.

Elder never decides on a strategy and executes tactics on the same timeframe. Given a monthly chart setup, I'm betting he'd step down to a weekly chart for entries & exits. I'm not saying his word is gospel, but I also typically use (at least) two timeframes. In this example I can't imagine waiting until the end of the month.

I choose a weekly EMA(21) to approximate a monthly EMA(5). Close enough. It seemed to do nice job below. At least this time. Obviously needs more testing, and obvioulsy other smarter folks than I will have their own ideas for an entry. But a weekly EMA(21) breach seems as good a place to start as any.

Thanks, Bruce!

You can't mix time frames within a single PCF, but you can create multiple conditions using different time frames and add them to the same EasyScan.

Yeah, that's what I thought. I knew that about the previous version, I just have to admit I've fallen behind on all the new (especially PCF) features in v17 and thought I'd ask to make sure. Never assume :)

Given that I'd like to try this on multiple timeframes anyway, I was anticipating using your parabolic PCF in one EasyScan to idenitfy candidates, and a separate EasyScan with multiple conditions and different timeframes to monitor the candidate's watchlist for possible entries. And, if it looks promising ... maybe another scan later for exits. Time will tell.

Thanks, Bruce!

Uncle, I completely agree that one would want a PCF to pop out possibilities. Any time between now and next spring, I'd imagine there will be some yummy ones. Please let us know if you think you've come up with something good. Thanks.

King,

A fews days ago Worden posted a short video with highlights of the Denver class. It's only 3 minutes but will give you a taste. Of course, all the customer's remarks are positive, or they wouldn't have posted the video. LOL!

But it will give you a peek at Michael Thompson, at least. He's awesome.

Again, attending the live class is highly recommended. I attend both days, but that's me. Others do, too, though I doubt many attend both days AND stay at the same hotel ... I'm guessing most come one day, or the other, as the planned content is the same.

Every day is a little different though, if only because the audience is different and asks different questions, which can take Michael down a different road. Also, since the market is open Friday and closed Saturday, you can't watch the market live on Saturday, but Michael usually has some recordings ready to demo on Saturdays.

Again, just me ... but if I had to drive 200 miles, I'd drive Thursday morning, check in Thursday afternoon, have some wine with dinner Thursday evening, get up early Friday morning, walk downstairs while everyone else is still driving and stake out a front row seat (I bring my laptop) then relax and enjoy breakfast inside the hotel. Then go back to the ballroom and catch Michael before all the other folks start rolling in and he gets busy. Best time to ask questions is before the crowd arrives. Then I chill Friday evening in the hotel and enjoy the Saturday session, filling in any gaps I missed Friday. Totally worth it. Best mini staycation I've ever taken.

As I mentioned before Michael is awesome, and he uses Worden's proprietary indicators on his charts everyday, so he could really help you out. If you don't want to wait until November, you could try email.

The only caveat is he's the one and only Worden trainer I'm aware of after Julia left, and then he got even busier when Worden added brokerage services. They even suspended live classes for a year or two. I'm not sure he answers emails these days (I could be wrong). I won't post his email online but Bruce or the customer support folks can pass it along if they think it's appropriate. You might be able to reach out to him regarding your BOP, TSV, and MS questions. Or, there's always November. :)

Hopefully folks will provide some feedback to your other post. 

Thank you for the video. I double-checked the class schedule and the next one within 200 miles of me is in November. I'll save up any questions I haven't been able to get answered by Support or on the Forums. Thanks again.

Sorry about the confusion Bruce. I should not have included Keltner Channels in the first place. My bad.

When you get a minute, can you post the final parabolic PCFs using closes instead of typical prices?

The salesman I buy my next sports car from thanks you. :)

As requested in the Average True Ranges topic, here it is again with the moving average of the close instead of typical price for the centerline (what I gave the first time).

TrueInRow(C > XAVG(C + 3 * ATR, 24), 2) = 2 AND ABS(XAVGC5 / XAVG(C + 3 * ATR, 24) - 1) <= .2 AND MIN(STOC1, 5) > 85

With the pullback being:

C < XAVG(C + 3 * ATR, 24) AND C < XAVGC5

And with pullback being true for the current bar and the original formula true for the previous bar.

C < XAVG(C + 3 * ATR, 24) AND C < XAVGC5 AND TrueInRow(C1 > XAVG(C1 + 3 * ATR1.1, 24), 2) = 2 AND ABS(XAVGC5.1 / XAVG(C1 + 3 * ATR1.1, 24) - 1) <= .2 AND MIN(STOC1.1.1, 5) > 85

That said, I have been thinking about the minimum drawndown requirement a bit and think I want to change it. The stochastic test as written checks for a minimum 15% drop from the high of each bar as a percentage of each bar.

MIN(STOC1.1.1, 5) > 85

What you probably really want is a 15% drop from the high of the 6 months down to the low after that high was made as a percentage of the high.

MINL5 >= .85 * H5 AND MINL4 >= .85 * MAXH2.4 AND MINL3 >= .85 * MAXH3.3 AND MINL2 >= 85 * MAXH4.2 AND L >= .85 * MAXH5.1

Which would make the first formula:

TrueInRow(C > XAVG(C + 3 * ATR, 24), 2) = 2 AND ABS(XAVGC5 / XAVG(C + 3 * ATR, 24) - 1) <= .2 AND MINL5 >= .85 * H5 AND MINL4 >= .85 * MAXH2.4 AND MINL3 >= .85 * MAXH3.3 AND MINL2 >= 85 * MAXH4.2 AND L >= .85 * MAXH5.1

And the pullback being true now with the first formula being true during the previous bar:

C < XAVG(C + 3 * ATR, 24) AND C < XAVGC5 AND TrueInRow(C1 > XAVG(C1 + 3 * ATR1.1, 24), 2) = 2 AND ABS(XAVGC5.1 / XAVG(C1 + 3 * ATR1.1, 24) - 1) <= .2 AND MINL5.1 >= .85 * H6 AND MINL4.1 >= .85 * MAXH2.5 AND MINL3.1 >= .85 * MAXH3.4 AND MINL2.1 >= 85 * MAXH4.3 AND L1 >= .85 * MAXH5.2

I think we're almost there.

First, I think your logic on the drawdown bit is spot on. I appreciate you thinking about it some more and improving it.

Being a programmer I'm totally embarassed by my lack of PCF literacy. I'm used to reading (and modifying) code written by other developers. Maybe I've relied on Bruce as a crutch too long.

I assume the "first formula" identifies the 3 criteria for a parabolic move, so we're cool there.

It's the "pullback being true..." PCF that I think might need changing, since my plan now is to step down to a lower timeframe and check the moving average there instead of the original timeframe.

So for this thread the pullback criteria is just a pullback inside the channel. We no longer need or want the moving average check. I'm going to do that in a separate time frame in a separate EasyScan.

So, after the stock goes parabolic and pulls back to close inside the channel, I'm putting it in a watchlist ... then running a separate scan on that list checking for an EMA(21) breach on a smaller timeframe.

Does that make sense? The good news is that Friday isn't all that far away. :)

It makes sense to me.

Cool. So how would the "pullback being true now with the first formula during the previous bar" PCF look without the 5-period moving average check?

We're so close. And I'm wearing out the scroll wheel on my mouse.

Please try the following.

C < XAVG(C + 3 * ATR, 24) AND TrueInRow(C1 > XAVG(C1 + 3 * ATR1.1, 24), 2) = 2 AND ABS(XAVGC5.1 / XAVG(C1 + 3 * ATR1.1, 24) - 1) <= .2 AND MINL5.1 >= .85 * H6 AND MINL4.1 >= .85 * MAXH2.5 AND MINL3.1 >= .85 * MAXH3.4 AND MINL2.1 >= 85 * MAXH4.3 AND L1 >= .85 * MAXH5.2

Thank you. For everything.

Uncle/Bruce: To understand better how formulas work, and to see if there were any current candidates, I set up a new EasyScan for US Stocks and copied the "first formula" into a formula condition. I let the default 'Is True, Now, Monthly' stay as it was. It didn't return anything.

I thought that it should at least return past instances, so I slowly increased "within last x bars" value until it was 360. I would think that there would be instances within the last 30 years, but it didn't return anything.

Because I only wanted to see the first half of the scenario, I didn't add the second condition.

Am I missing another condition in the setup? I've only set up a few scans, so I probably missed something. Any input would be appreciated.

On a different note, I see that Bruce moved the other topic to the other sub-forum. Maybe it will find an audience there. Thank you, Bruce.

Yeah, I suggested Bruce move the other topic. Didn't mean to speak on your behalf - and I didn't think Bruce would be that quick - I figured he'd wait a while and let you see my suggestion. Fyi, I just posted 3 links there to good short reads by Martha Stokes, CMT. Smart lady. Check it out.

I admit I haven't tested the parabolic PCFs yet; usually get my hands dirty on the weekends. TC2000 has 500 bars of data, so you should be able to test the PCF against the Apple 2012 example above where it went parabolic early in the year and finally went Wile E. Coyote in October.

There's a way to plot a spike or other visual cues on a chart to help visually "backtest" with your eyeballs on a chart. Bruce can explain it better than I.

Thanks for testing this out!

Requiring max drawdown to be less than 15% over the most recent 6 months is generally the killer. Without this requirement, you will get quite a few results.

TrueInRow(C > XAVG(C + 3 * ATR, 24), 2) = 2 AND ABS(XAVGC5 / XAVG(C + 3 * ATR, 24) - 1) <= .2

It got lots of results. I'll study them and study the change. This gives me plenty to experiment with.

One reason I'd like to know this (and I bet plenty of other people do, too), is for existing holdings. With the market behavior over the last 18 months, I own several stocks that have appreciated a fair amount, and which could go parabolic if one scenario plays out. If nothing else, this scan would indicate sell timing.

Thanks, Bruce, and thanks to Uncle for introducing the topic!

Bruce, that's great to know. Thanks!

Funny, I just re-read the article and he actually says "Minimum drawdown from high to close is less than 15% in the past six months" in the bullet point listing the criteria. Either the editor didn't know squat about trading or they didn't proofread the article.

Elder gets it right in the next paragraph ("... it has not experienced a closing drawdown of more than 15% in the past six months").

Just another reminder to not believe everything you read online. :)

What I usually do when testing scans based on ideas I steal from someone else is to adjust the parameters for one criteria at a time while leaving the others alone and see how that impacts the number of results returned. Gives me a feel for how I want to proceed. Everyone has their own thing and I guarantee no two people who read this are going to wind up using this scan exactly the same way.

Yeah, that criteria may need to be relaxed, or eliminated altogether.

It is still really restrictive if we assume the "drawdown" is to the lowest close after the high.

TrueInRow(C > XAVG(C + 3 * ATR, 24), 2) = 2 AND ABS(XAVGC5 / XAVG(C + 3 * ATR, 24) - 1) <= .2 AND MINC5 >= .85 * H5 AND MINC4 >= .85 * MAXH2.4 AND MINC3 >= .85 * MAXH3.3 AND MINC2 >= 85 * MAXH4.2 AND C >= .85 * MAXH5.1

It is only when we let the "drawdown" be to the close of the same bar that it becomes a lot less restrictive.

TrueInRow(C > XAVG(C + 3 * ATR, 24), 2) = 2 AND ABS(XAVGC5 / XAVG(C + 3 * ATR, 24) - 1) <= .2 AND TrueInRow(C >= .85 * H, 6) = 6

I admit I haven't given the drawdown component of this much thought up to now. 

Given the language used in the article, and the potentially sloppy editing, it maybe time to think like a trader and not a computer, an author, or an editor :)

I pasted the excerpt of the article with the only two references to drawdown below. When I first read this (chugging coffee and bookmarking it for later study) I thought he meant a maximum drawdown of 15% from its highest point in its last six months to any point during the last six months. I guess on a monthly chart that would make it the lowest low of the last six candles. That's probably the most restrictive interpretation, but that was my thinking at the time. 

I don't hold long-term (probably should but that's another story) I would look at drawdown from the point of view of the highest high while I was holding to the current price, whenever that was. I wouldn't look at justthis month's drawdown, but that's me. If I lost 10% in March and 5% in April, I've lost 15% rather than 5%.

Anyways, I'm not going to try and get in Elder's head, but I think the drawdown requirement might have been intentionally restrictive because the article was about "parabolic" moves. Everyone's definition of a parabola is going to be different (except for mathematicians, LOL) but my point is that when I visualize a parabolic stock chart I see and uptrend that gradually and then exponentially accelerates, without a lot of choppiness or volatility (or drawdowns). To me the that's the essence of a parabolic move. Nobody sells, so there's almost no drawdown, just more and more lemmings pile in until the inevitable.

I'm not a CMT, and not even very smart, but to me "parabolic" and "overextended" or "overbought" are two separate concepts. Again, just me. To me, parabolic is (almost) as rare as the Cubs winning the World Series.

So, I think Elder was right. Since it was a MarketWatch.com article and not a Stocks & Commoditites article, we might never know what he meant exactly. An S&C article would include actual code that we could port over to TC2000.

And I think Bruce was awesome for identifying the drawdown criteria as the most restrictive of the three. I think the garden variety overbought/overextended stocks will have more volatilty but being above the 3-ATR threshold for at least two candles and having the EMA(5) within spitting distance of the 3-ATR threshold are more important for scanning purposes.

That took waaaay too long. I'm too tired to proofread this one. Taking a short break; going for a walk. Sorry about any typos. :)

From the original article:

• Two closes outside of a three-standard deviation channel of a 24-month Average True Range (ATR).
• Five-month moving average is within 20% of the three standard deviation of the 24 month ATR (monthly 3-ATR channel).
• Minimum drawdown from high to close is less than 15% in the past six months.

I'm back. I see Elder mentions close when decribing drawdowns, and that appears to be what Bruce used, so that's fine.

I'd use the lowest six month low, but again, I'm not a long term holder, and this is probably why ... can't imagine not checking the price for an entire month. I'd look every day. LOL. Don't have the personality for it. Probably should work on that.

So fine! We have two great scan PCFs, courtesy of Bruce, based on an excellent premise, courtesy of UnclePennybags.

From looking at a hundred charts, including those of big-moving stocks that don’t qualify in these two scans (with, and without the ‘drawdown’), we see that another concept would be to scan for stocks that recently closed above their 5 EMA and ATR centerline for two three consecutive months. ‘Recently’ might equal ‘in the last 6 months.’ There could possibly be a third component that they are approaching the upper channel boundary (don't know if that matters, or if it's possible).

I ask Uncle and Bruce to weigh in and fine-tune this concept.

The point is to find (via a relatively short list) which stocks are on the way to breaking out to a parabolic move. All of the stocks in the ‘first formula’ spent some time testing their 5 EMA and ATR centerline, then closed above it for two three consecutive months. Which of them might go parabolic is not discernible. But having a manageable list of candidates to check weekly based on Uncle's 2nd chart would be very interesting, don’t you think? (Especially in our current market, no?)

BTW, in setting up my scans, I limited volume 50DMA to > 500K. It might be too restrictive, but for one of my future purposes, it’s a must. YMMV.

Thanks again to Bruce and UnclePennybags.

Back in a minute. Copying above text to Word to remove all those &lsquo and &rsquo. I've only had one cup of coffee so far and I'm having trouble reading this :)

Darn. Sorry 'bout that. They're apostrophes and single quotes.

A condition formula for price being above both its 5 EMA and 24 ATR exponential centerline for at least 2 consecutive months within the last 6 months can be written as follows.

CountTrue(TrueInRow(C > XAVGC5 AND C > XAVGC24, 2) = 2, 5) > 0

If you need it to be for at least 3 consecutive months during the last 6 months.

CountTrue(TrueInRow(C > XAVGC5 AND C > XAVGC24, 3) = 3, 4) > 0

Approaching the upper ATR band would depend on what you mean. It could just mean that price is moving up.

C > C1

Or that price is getting closer to the upper band.

XAVG(C + 3 * ATR, 24) - C < XAVG(C1 + 3 * ATR1.1, 24) - C1

But you could also check for price to be say above 2 multiples of ATR instead of 3.

C > XAVG(C + 2 * ATR, 24)